Ecological and evolutionary implications of spatial heterogeneity during the off-season for a wild plant pathogen

While recent studies have elucidated many of the factors driving parasite dynamics during the growing season, the ecological and evolutionary dynamics during the off-season (i.e. the period between growing seasons) remain largely unexplored. We combined large-scale surveys and detailed experiments to investigate the overwintering success of the specialist plant pathogen Podosphaera plantaginis on its patchily distributed host plant Plantago lanceolata in the Åland Islands. Twelve years of epidemiological data establish the off-season as a crucial stage in pathogen metapopulation dynamics, with c. 40% of the populations going extinct during the off-season. At the end of the growing season, we observed environmentally mediated variation in the production of resting structures, with major consequences for spring infection at spatial scales ranging from single individuals to populations within a metapopulation. Reciprocal transplant experiments further demonstrated that pathogen population of origin and overwintering site jointly shaped infection intensity in spring, with a weak signal of parasite adaptation to the local off-season environment. We conclude that environmentally mediated changes in the distribution and evolution of parasites during the off-season are crucial for our understanding of host–parasite dynamics, with applied implications for combating parasites and diseases in agriculture, wildlife and human disease systems.

. Factors that affect the fraction of infected leaves with resting structures of the pathogen Podosphaera plantaginis in its host populations in the Åland Islands, southwestern Finland. Shown are the results from year-specific spatial Bayesian models using LAPLACE approximation.
Estimates in bold are significant as based on a decrease in DIC. The spatial pattern of resting spore production is shown in Fig. 4 in the main manuscript.     southwestern Finland. Shown are the results from year-specific spatial Bayesian models using LAPLACE approximation. The spatial pattern of resting spore production is visualized in Fig. 4 S2).
Notes S1. A trial experiment on overwintering survival using indoor and outdoor overwintering sites

Aims
We carried out a trial overwintering experiment to simultaneously i) provide a first demonstration that resting structures are able to infect plants in spring in this pathosystem, and ii) assess the impact of winter conditions on the survival of resting structures collected from different populations.

Materials and methods
We shown that infection develops well in these bags, and spores cannot leave or enter (Laine, 2011).
Plants were scored on 21 June for the presence of powdery mildew infection.

Results
Environmental conditions during the off-season had a major impact on viability and spring germination of the resting structures. The experiment revealed that none of the resting structures stored indoors were able to infect caged plants in spring, even though they were stored at a wide range of temperatures (-10 to +10 °C; Fig. S5). This is in striking contrast with an average infection percentage of 34% (11 out of 32 caged plants) when resting structures were stored overwinter in natural populations (Fig. S5). There is also an indication that off-season survival in the field varied among the overwintering sites (Fig. S5).
Methods S1. A detailed description of the statistical methods.
To analyse the impact of environmental and spatial factors on the spatial pattern of winter extinction, July abundance and the proportion of infected leaves with resting structures, we fitted a Bayesian spatial model using the integrated nested Laplace approximation (Cameletti et al., 2012) as implemented in the package INLA (Rue et al., 2009;Lindgren et al., 2011) in R version 2.15.1 (R Core Team, 2012). The advantage of this method is that it efficiently and accurately estimates both covariates and the spatial range of autocorrelation (as based on Euclidean distance between populations). For both overwintering survival and resting structure formation, we included the environmental variables distance to shore, plant dryness, patch shadow, habitat openness, July rainfall, August rainfall and population age (i.e. how many years ago the pathogen population had been established by colonization, with a maximum value of 5) and the spatial factors host plant coverage, road presence and host plant spatial connectivity as explanatory covariates. The average rainfall in July and August was estimated separately for each population using detailed radarmeasured rainfall data. To reduce the number of covariates in the model we pre-selected the covariates to be included using a linear / logistic model and the function stepAIC with the option 'backwards' (package MASS). Significance of the explanatory variables was then assessed based on the deviance information criterion (DIC) in the spatial Bayesian model.
To analyse the experimental data, we used the framework of generalized linear mixed-effects models (Littell et al., 2006). All models were fitted with procedure GLIMMIX in SAS 9.3. For binomial data, we assumed a binomial distribution with a logit link. For models with multiple interactions, we used the principle of backwards stepwise model simplification to arrive at a minimum adequate model, where variables were retained when p<0.1 (Crawley, 2007).
Significance for fixed and random effects was assessed using F-tests and log-likelihood ratio tests, respectively (Littell et al., 2006).
In the overwintering experiment we aimed to investigate the impact of pathogen population of origin and overwintering site on overwintering success. To analyse this experiment, we modelled the response variables Infection (0/1) and would capture variation in disease intensity resulting from infection by resting structures that were stored in sympatry or allopatry (i.e. resting structures that were overwintered in the location from which they were collected or in a non-local location, respectively).
Finally, we investigated the relationship between the presence of resting structures and offseason survival at two spatial scales. At the plant level, infection (0/1) and disease intensity (number of infected leaves / total number of leaves) in spring were modelled as a function of the number of leaves with resting structures. The pathogen population was used as random factor to account for variation among populations in overwintering success. At the patch level, off-season survival and July abundance were modelled as a function of the fraction of infected leaves with resting structures in the previous autumn.